AI “lead” may matter less than expected, IMD professor says, citing OpenAI and Anthropic’s edge as “very perishable”

The debate over who is winning the artificial intelligence arms race is increasingly shifting from raw technical capability to how durable that advantage will be. In comments highlighted by Yahoo Finance, Howard Yu, LEGO Professor of Management and Innovation at IMD, characterized the competitive edge of top U.S. AI labs as “very perishable,” suggesting that leadership in cutting-edge models may not translate into long-lasting dominance.

Yu’s remarks were framed around the relative positions of OpenAI and Anthropic, both of which are often viewed as among the most advanced developers of frontier AI systems. The professor’s central point was not that these companies lack strengths, but that technological superiority can diminish faster than markets and strategists may assume as rivals improve and as new products and deployment approaches spread across the industry.

The interview also used a consumer-business analogy to make the case that having the best technology is not the only path to commercial success. In Yu’s formulation, organizations can reach major, high-volume goals without necessarily needing to drive the “Ferrari,” a reference to maintaining an uncompromising, top-of-the-line approach when a cost-effective route could be sufficient.

While the comments discussed AI competition broadly, the analogy directly connects the frontier-model race to real-world scaling. The implication is that retailers, employers, and other adopters may prioritize practical performance, integration, and total cost over marginal gains that only a small share of users would notice.

For Walmart, the relevance is more about the translation of AI capability into everyday operations than about winning an abstract benchmark. Retailers and other high-throughput businesses typically care about how AI tools affect customer experiences, merchandising accuracy, supply-chain efficiency, and workforce productivity. In this context, the “perishable lead” idea suggests that adoption strategies should account for rapid improvement across vendors, not just the current state of model quality.

The article did not provide specific technical metrics comparing OpenAI, Anthropic, or any other competitor, and it did not lay out a timeline for how quickly an advantage might fade. It also did not quantify how long leading models might retain a measurable edge in deployment performance, such as latency, reliability, cost per request, or enterprise integration time.

The remarks were positioned as a managerial insight into innovation and competition rather than a company-specific forecast. That matters because it underscores a shift in how executives may evaluate the AI landscape: not only whether a provider is best today, but whether today’s best is likely to remain the best after competitors adjust, pricing pressures increase, and software layers standardize.